Feature Selection for Object-Based Classification of High-Resolution Remote Sensing Images Based on the Combination of a Genetic Algorithm and Tabu Search

Table 5

List of features selected by each method.

Data	Method	Features
		Spectral	Texture	Geometry
WorldView-2	GATS	Mean layer 1/4, Brightness,	GLCM mean layer 3 (0°),	—
		Ratio layer 1, Intensity	GLCM ang. 2nd moment (135°)
	GA	NDWI, NDVI,	GLCM homogeneity layer 1 (45°),	—
		Ratio layer 1/2/4	GLCM homogeneity layer 3 (45°)
	TS	NDWI, NDVI,	—	—
		Mean layer 2/4,
		Ratio layer 1/2/4
	ReliefF	NDWI, NDVI,	—	Elliptic Fit
		Mean layer 4, Brightness,
		Ratio layer 1/4
QuickBird	GATS	NDWI, NDVI,	GLCM ang. 2nd all dir.,	—
		Mean layer 2,	GLCM ang. 2nd moment
		Ratio layer 1, Intensity	layer 1 (45°)
	GA	NDWI, NDVI,	—	—
		Mean layer 3/4,
		Ratio layer 2/3, Saturation
	TS	NDWI, NDVI,	—	—
		Brightness,
		Ratio layer 1/2/3/4
	ReliefF	NDWI, NDVI,	—	—
		Mean layer 3, Saturation,
		Ratio layer 3/4, Hue